Update app.py

app.py

@@ -1,31 +1,48 @@
 # -*- encoding: utf-8 -*-
 import gradio as gr
-from gradio.components import label
-import requests
-import json
+import random
+import time
+from concurrent.futures import ThreadPoolExecutor
+import asyncio
+from functools import partial
+from volcengine.maas import MaasService, MaasException, ChatRole
+maas = MaasService('maas-api.ml-platform-cn-beijing.volces.com', 'cn-beijing', connection_timeout=24000, socket_timeout=24000)
+maas.set_ak("AKLTY2ZjOWM0NmFlNTEwNDBhM2EyYTg4OTgzYTUyYTc2NjU")
 
-def llm_infer(messages, character_name, character_info, user_name, user_info):
-    url='https://search.bytedance.net/gpt/openapi/online/v2/crawl?ak=paQxYUZ0TcbuBSFPgSbwR5QBdizjW1Ut'
-    headers={
-        'Content-Type': 'application/json'
-    }
-    data={
-        "messages": messages,
-        "character_profile":{
-        "character_name":character_name,
-        "character_info":character_info,
-        "user_name":user_name,
-        "user_info":user_info
-    },
-    "model": "Baichuan-NPC-Turbo",
-    "stream": False
-    }
-    ret=requests.post(url=url, headers=headers, data=json.dumps(data))
-    
-    return ret.json()['choices'][0]['message']['content']
+maas.set_sk("T0Rjd01HUTFaVEF3Tm1FME5EWXhNRGhpTURreE1ETTROalkwTnpreU1UVQ==")
+
+def llm_infer(endpoint_id, messages, agent_network_output=True):
+    req = {
+        "model":{
+           "endpoint_id":endpoint_id
+        },
+        "parameters": {
+            "max_prompt_tokens": 3800,  # 最大prompt，自动截断前面的输入。0-不生效
+            "max_new_tokens": 0,  # 输出文本的最大tokens限制。0-不生效
+            "min_new_tokens": 0,  # 输出文本的最小tokens限制。0-不生效
+            "temperature":1,  # 用于控制生成文本的随机性和创造性，Temperature值越大随机性越大。取值范围0~1
+            "top_p": 0.7,  # 用于控制输出tokens的多样性，TopP值越大输出的tokens类型越丰富。取值范围0~1
+            "top_k": 0,  # 选择预测值最大的k个token进行采样，取值范围0-1000。0-不生效
+        },
+            "messages": messages,
+        }
+    while True:
+        try:
+            resp = maas.chat(req)
+            break
+        except MaasException as e:
+
+            print(e)
+            continue
 
-def respond(message, history, character_name, character_info, user_name, user_info):
+    return resp['choice']['message']['content']
+
+def respond(message, history, ep: gr.TextArea, sp:gr.TextArea):
     messages = [
+        {
+        'role':'system',
+        'content':sp
+        },
         ]
     for part in history:
         messages.extend(
@@ -45,35 +62,61 @@ def respond(message, history, character_name, character_info, user_name, user_in
         'content':message
     })
     print(messages)
-    ret=llm_infer(messages, character_name, character_info, user_name, user_info)
+    ret=llm_infer(ep.strip(), messages)
     history.append([
         message,
         ret
     ])
     #if len(history) % 2 == 0:
     return "", history
-CSS ="""
-.contain { display: flex; flex-direction: column; }
-#component-0 { height: 100%; }
-#chatbot { flex-grow: 1; }
-"""
 
-with gr.Blocks(fill_height=True, css=CSS) as demo:
-    with gr.Row(equal_height=True): 
-        with gr.Column(scale=2):
-            chatbot = gr.Chatbot(elem_id="chatbot")
-            chatbot.height=550
+
+with gr.Blocks(fill_height=True).queue(default_concurrency_limit=10) as demo:
+    with gr.Row(): 
+        with gr.Column():
+            chatbot = gr.Chatbot()
+            chatbot.height = 500
         with gr.Column():
-            character_name=gr.TextArea(label="角色名", max_lines=1, lines=1, value='雷雷')
-            character_info=gr.TextArea(label="角色信息", max_lines=3, lines=1, value='小学生，在人大附小上学')
-            user_name=gr.TextArea(label="用户名", lines=1, max_lines=1, value='明明')
-            user_info=gr.TextArea(label='用户信息', lines=1, max_lines=3, value='明明是雷雷的哥哥，是个初中生')
+            gr.Label("input sp/ep of bot") # 增加些描述
+            sp=gr.TextArea(label="你是...", lines=1, max_lines=5,value='')
+            ep=gr.TextArea(label='ep1', lines=1, max_lines=1, value='')
             
             
     with gr.Row():
         chat_input = gr.Textbox(label="input")
+    
+    def handle_click(chatbot, ep, sp, evt:gr.SelectData):
+        messages = [
+            {
+            'role':'system',
+            'content':sp
+            },
+        ]
+        index = evt.index[0]
+        print(index)
+        for part in chatbot[:index]:
+            messages.extend(
+                [
+                    {
+                        'role':'user',
+                        'content':part[0]
+                    },
+                    {
+                        'role':'assistant',
+                        'content':part[1]
+                    }
+                ]
+            )
+        ret = llm_infer(ep, messages)
+        print(ret)
+        print(chatbot[:index] + [[chatbot[index][0], ret]] + chatbot[index + 1:])
+        return chatbot[:index] + [[chatbot[index][0], ret]] + chatbot[index + 1:]
    
-    chat_input.submit(respond, [chat_input,chatbot, character_name, character_info, user_name, user_info],[chat_input, chatbot])                    
+    chat_input.submit(respond, [chat_input,chatbot, ep, sp ],[chat_input, chatbot])                    
+    #chatbot.select(handle_click, [chatbot, ep, sp],[chatbot],show_progress='hidden')
     clear = gr.ClearButton([chat_input, chatbot],)
     
-demo.launch()
+    
+    
+    
+demo.launch(share=True)